minimal-run-and-audit

Standardized execution and audit reporting for deep learning repository reproduction runs.

- Captures evidence from smoke tests, inference runs, and evaluation commands; writes normalized outputs to repro_outputs/ with patch tracking when repository files change

- Generates SCIENTIFIC_CHANGELOG.md to document changes affecting evaluation, preprocessing, or metrics, and COMPARABILITY_REPORT.md to assess alignment with README and paper baselines

- Applies only after a reproduction target and setup plan exist; does not handle initial repo intake, training execution, or target selection

- Distinguishes between verified, partial, and blocked execution states; refuses to hide changes that alter scientific meaning

minimal-run-and-audit

Use the shared operating principles in
../../references/agent-operating-principles.md; this skill should make run
evidence auditable without turning every command into a rigid protocol.

When to apply

- After a reproduction target and setup plan exist.

- When the main skill needs execution evidence and normalized outputs.

- When a smoke test, documented inference run, documented evaluation run, or other short non-training verification is appropriate.

- When the user already knows what command should be attempted and wants execution plus reporting only.

When not to apply

- During initial repo scanning.

- When environment or assets are still undefined enough to make execution meaningless.

- When the task is a literature lookup rather than repository execution.

- When the user is still deciding which reproduction target should count as the main run.